maximum number in a Single precision floating number

[JCNT]

Hi everyone,

I would like to know which is the maximum positive and negative number I can write in a SGL numeric control. I read the Numeric data types table  in the LabVIEW help and I am not sure if the numbers in the column "approximate range on disk" are the maximun and minimum ones because I wrote in a DBL numeric control the number 6216825840 (which is less than the maximum on the table 3.40e+38) and I displayed it in a SGL numeric indicator, and the number changed. 

The problem is that I am storing some data in the numeric format and  I choose the SGL format because it uses less space in the disk, however I have the number 6216825840 which changes if it is storaged in this format. I know that I should change to DBL, but I am doing an array with all my data and if I change one by BDL the other ones change to DBL and will use more space in the disk when I store them.

I attach a picture of this problem.

Thanks in advance.

JC

[Donald]

---

JCNT wrote:

I would like to know which is the maximum positive and negative number I can write in a SGL numeric control. ....

I wrote in a DBL numeric control the number 6216825840 (which is less than the maximum on the table 3.40e+38) and I displayed it in a SGL numeric indicator, and the number changed. 

---

This is a precision problem/limitation, not a max/min problem.

SGL is limited to ~7 digits precision.
With DBL, you have ~15 digits precision.

[JCNT]

Hi Donald,

Thanks for reply,

You mean that I could  just write until the number 9999999?

JC

[Donald]

---

@JCNT wrote:

You mean that I could  just write until the number 9999999?

---

No.  You are not limited to 7 digit numbers ... as shown by the 10 digit SGL 6216825856 in your JPG.

But a SGL cannot give you precision beyond the 7th digit: 6216825xxx

Message Edited by Donald on 06-15-2007 03:40 PM

[altenbach]

And since the numbers are stored in binary, the resolution in decimal digits is just approximate.

SGL has a 23 bit mantissa (http://en.wikipedia.org/wiki/Single_precision), so you cannot do better than 2^24 for integers.

Your number does not even fit into a U32, which has all 32 bits reserved for the integer part. 😮

(DBL has a 52 bit mantissa.)

[altenbach]

In particular, you can create a simple test program that shows you that the highest good integer is 16777216 for SGL, because 16777216+1=16777216.

For deep reasons, this is the same as 2^24 as mentioned above. 🙂

Here's a quick code that stops the loop once i = i+1, then displays i (the fist integer that fails the test, thus is the largest good integer for SGL. 🙂

Message Edited by altenbach on 06-15-2007 01:33 PM

[JCNT]

Hi Donald and altenbach,

Thanks so much both of you for your help, now I understand what you mean. So, I don't have any choice and I must store my data in DBL format.

Thanks again and I'm really glad that there are people like you who help people with these problems.

Regards from Ecuador!

JC